Installation of LEGION

March 31, 2025 · View on GitHub

1. Prerequisites

Tip

You can skip this section if you have installed all required packages.

1.1 The code is installed and tested on Ubuntu 22.04 LTS, make sure your OS is Ubuntu 20.04 or 22.04

1.2 We use miniconda to manage the virtual environment, make sure you have installed miniconda following: https://docs.anaconda.com/miniconda/

These four commands quickly and quietly install the latest 64-bit version of the installer and then clean up after themselves. To install a different version or architecture of Miniconda for Linux, change the name of the .sh installer in the wget command.

mkdir -p ~/miniconda3
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh -O ~/miniconda3/miniconda.sh
bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
rm -rf ~/miniconda3/miniconda.sh

After installing, initialize your newly-installed Miniconda. The following commands initialize for bash and zsh shells:

~/miniconda3/bin/conda init bash
~/miniconda3/bin/conda init zsh

1.3 We use Mujoco200 to build RL environment

To install Mujoco200, following: https://www.roboti.us/download.html, download mujoco200 linux, unzip the file and place at your/home/path/to/.mujoco/ with the folder name mujoco200. Namely, the full path should be something like /home/ubuntu/.mujoco/mujoco200. After that, add following command line (modify according to your own path) in your .bashrc file:

export  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/your/.mujoco/mujoco200/bin

2. Installation of LEGION dependencies.

Here we introduce how to establish the legion dependencies.

2.1 build virtual env:

Open a terminal:

conda create -n legion python=3.7

conda activate legion

pip install dm_control mujoco-py==2.0.2.8 cython==0.29.33 protobuf==3.20.0 gym==0.20.0

2.2 Download and install LEGION dependencies:

Install legion repository dependencies

(optinal) git clone https://github.com/Ghiara/LEGION.git

cd LEGION && conda activate legion && pip install -r requirements/dev.txt

Install modified metaworld environments under source folder

cd source

git clone https://github.com/Ghiara/Metaworld-KUKA-IIWA-R800.git

cd Metaworld-KUKA-IIWA-R800 && pip install -e . && cd ..

Install Bayesian Nonparametric Models library (bnpy) under source folder

git clone https://github.com/bnpy/bnpy.git

cd bnpy && pip install -e . && cd ..

Install mtenv modified environment manager

git clone https://github.com/Ghiara/mtenv.git

cd mtenv && pip install -e . && cd ..

Important

Before you run the code, change the task_encoder_cfg.path_to_load_from in the config/agent/components/continuouslearning_multitask.yaml with your local repository path.

3. Error catching

Warning

TypeError: Descriptors cannot not be created directly. If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.If you cannot immediately regenerate your protos, some other possible workarounds are: 1. Downgrade the protobuf package to 3.20.x or lower.2. Set PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python (but this will use pure-Python parsing and will be much slower).

pip install protobuf==3.20.0

Warning

Failed import mujoco-py:

try another version of mujoco-py, for example:

pip install mujoco-py==2.0.2.10

Warning

ImportError: /home/.../miniconda3/envs/legion/bin/../lib/libstdc++.so.6: version `GLIBCXX_3.4.30' not found (required by /lib/x86_64-linux-gnu/libLLVM-13.so.1)

Refer this page: https://bcourses.berkeley.edu/courses/1478831/pages/glibcxx-missing

cd /home/to/your/anaconda3/envs/legion/lib
mv libstdc++.so.6 libstdc++.so.6.old
ln -s /usr/lib/x86_64-linux-gnu/libstdc++.so.6 libstdc++.so.6